Researcher Sophie Leroy studied what goes on when people leave one task for another before the first one feels done. She discovered that some of your focus remains on the previous task, causing the next task to rely on what’s left of your attention. This is called attention residue. The change on your calendar was immediate, but the change in your mind takes time.

Switching between tasks has a hidden cost. Today you build a fix for it: a 90-second reset you run between tasks. You come out of it having closed the task you just left and knowing the exact first thing to do next.

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction, looking to gain traction on their complex lives.

Here are some recent popular builds you may have missed:

➡ A Claude project that pinpoints why you keep dodging one task, then hands you a first step small enough to start before you talk yourself out of it.

➡ A Notion and Claude setup that files every note and half-thought you dump in each Monday and shows you which projects have gone cold.

➡ A no-code Claude planner that gives your hardest work to your sharpest hours and keeps your drained hours light.

➡ A single prompt that runs your weekly review in ten minutes and gives you a 30-second first step for each priority.

Below, I’ll show you how to build the 4-step reset and walk you through a switch from start to finish. By the end, you’ll be able to leave one task and drop straight into the next one at full focus, instead of losing your first 20 minutes to a task your mind should have already left behind.

Read more

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction, looking to gain traction on their complex lives.

Here are some recent popular builds you may have missed:

➡ A ten-minute weekly review you run with AI that turns next week’s priorities into first steps small enough to actually start, so the list doesn’t just sit there.

➡ An agent that writes tomorrow’s ranked to-do list the night before, so you know your first task before you sit down.

➡ A Notion and Claude setup that files every note, link, and half-thought you dump in into the right place each Monday, and shows you which projects have gone cold.

➡ One rule for deciding which AI tools to trust with your work, and which to keep away from your email, files, and bank.

➡ A Claude project that finds the reason you keep dodging a task, then hands you a first step small enough to start before you talk yourself out of it.

Start with a decision you make often. Like you’re choosing between two directions for a project. You give an AI both options and ask it to compare them.

It does a good job. It lays out the trade-offs, pressure-tests each side, and flags a risk in one of them you hadn’t spotted. Then it recommends one of the two directions and explains why. The topic looks settled.

You’ll be tempted to take that recommendation and move on. That’s the moment where you need to slow down a little.

Subscribe now

A decision like this has two separate parts. First, someone has to gather the evidence: line up the trade-offs, test each direction, find the risks. Second, someone has to sit with that evidence and decide which direction to choose. AI is good at that first part. Gathering evidence takes time and effort, and handing it to AI genuinely helps. Deciding which direction to head in asks something different of you. You have to weigh the evidence yourself and commit to one choice, knowing you’re accountable for it. When AI has already recommended a direction, most people agree with it instead of weighing the options themselves. The answer is already written, and thinking critically is harder than accepting it.

Researchers have a name for this and have measured how often it happens. In a study of 9,000 people across nine countries, participants shown an AI recommendation often switched away from their own correct answer to match it, even when their first answer was correct. The study also found people switched toward the AI more often when it was described as highly tested and accurate. The more authoritative the recommendation seemed, the more readily people abandoned their own judgment.

The same study found something less obvious about who defers. People with no experience with AI tended to distrust it, so they stuck with their own decision. People with deep expertise leaned on the AI when it had proven accurate before and doubted it when it hadn’t. The people who over-relied on it sat in the middle: familiar enough with AI to trust what it produced, but not experienced enough to notice when it gave inaccurate outputs. The researchers call this middle zone a beginner’s bubble.

That middle zone describes how most of us approach most decisions at work. You may know your field well and still not have a firm answer to the exact question in front of you. This is where you’re most likely to adopt the AI’s answer without examining it.

Subscribe now

This leads to which part of the decision-making you want to protect. Let AI gather the evidence. When it comes to choosing a direction, read the evidence and decide what you think before you look at the AI’s recommendation, and be ready to stand by your own reasoning when your answer and its answer differ. When you skip that step, the AI’s recommendation becomes the decision, and you approve it. It feels like deciding, but if someone asks why you chose that direction, the reasoning was the AI’s and you are left explaining a choice you didn’t actually reason through.

There’s a second cost here that compounds over time. Another study of nearly 600 people tracked what happens when people repeatedly let AI do their thinking for them. The more they let AI decide, the less motivated they were to work through the next problem themselves, and the more readily they let it decide that one too. It’s a slope. The concerning part is that none of this is visible while it’s happening. In that study, letting AI manage the thought process felt just as useful and efficient to people as letting it handle a simple chore. Both gave the same satisfying dopamine hit of getting a task done. That is why the cost of handing over your thinking is so easy to miss: nothing about the experience warns you in the moment. The task gets done, which feels fine. The catch is that each time you accept the recommendation instead of judging the evidence, you get a little more practiced at deferring and a little less practiced at thinking critically.

Once you know where that risk is greatest, you can sort your tasks by it. Automate the evidence-gathering on every decision you can. When the result has to be a judgment you are accountable for, decide it yourself and use AI to strengthen your thinking rather than replace it. Have it lay out the case, challenge your reasoning, and surface what you missed. Then decide what you think before you read its recommendation. You still move fast on everything around the decision, but you’re the one making the decision itself.

Knowing which tasks to keep is half the battle. The other half is the friction that’s pure waste, the stuck feeling in the gap between tasks, where you’ve finished one thing and can’t get into the next. Sunday’s build is an AI reset ritual for that stuck stretch between tasks. Rest between tasks is good and worth protecting. This is for the different feeling of being unable to start the next thing even when you want to, so the changeover stops costing you time you never meant to lose. Read it here:

Subscribe now

Altman pointed to the economy as the reason why OpenAI’s disruption didn’t show up on the timeline they predicted. ‘The economy just has so much inertia,’ he said. ‘We’ve all been too ambitious on timelines. Even with this incredible technology, society and the economy will adapt more slowly. People keep doing the same things, buying from the same company, wanting to use their tools the same way,’ he said. ‘I think this is actually a positive in many ways, and it’s going to make this big transition go smoother and slower. I’m grateful for it.’

On the macroeconomic level, he’s right. A new tool doesn’t rewire a supply chain in a quarter. Companies have contracts, staff, and habits, plus a hundred reasons not to change this year. The technology can sprint while the market walks behind it. His point about slow adoption is fair.

Whether this is more PR strategy from Altman and OpenAI to counter the employment doomsday narrative they’ve been pushing for the past 4 years as they try to push through this IPO is debatable. However, you can’t deny that a smaller version of this appears in your own workday. This one you can influence, though.

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction and looking to gain traction in their complex lives.

Here are some recent popular builds you may have missed:

➡ A no-code energy-aware planner that gives your hardest work to your sharpest hours and keeps your drained hours light, so you stop scheduling your day as if morning you and 3 pm you are the same person.

➡ An AI system that finds the reason you keep dodging a task, then hands you a first step small enough to start before you talk yourself out of it.

➡ A Notion and Claude second brain that sorts everything you capture into place each Monday and tells you which projects have gone quiet and what you’ve been avoiding.

➡ An AI decision agent that writes tomorrow’s ranked to-do list the night before, so you know your first task before you sit down.

➡ A ten-minute weekly review you run with AI that turns next week’s priorities into first steps small enough to actually start, instead of handing you a list you’ll never start.

You’ve probably already used AI to help you complete individual tasks more quickly. It can draft emails, locate buried files, pull data, and provide a first version of reports, saving you time. However, despite these efficiencies, you still find yourself just as busy. You get through each task faster, yet by 5 PM, you feel just as exhausted.

That exhaustion is your own version of the economic inertia Altman describes, except here it isn’t the economy’s fault, and it isn’t yours. The friction that drains your day never comes from the tasks themselves. Your work becomes difficult in two distinct ways.

Subscribe now

One kind is mechanical. The file is buried in the wrong folder. You’re copying numbers from one app into another by hand. You’re rebuilding the same context for the fifth time. These are chores, wasted effort, and AI should take them off your hands. It can now.

The other type of task is more challenging because it requires understanding its purpose. You’re in the process of figuring out your thoughts, deciding what matters, and determining what to eliminate. You need to make sure you haven’t framed the entire project incorrectly before wasting a day working on the wrong solution. AI can help with this too; just give it the initial mess, and it will deliver a polished answer.

AI can handle both types of tasks. For the more mechanical tasks, you can delegate them entirely to AI. For more conceptual tasks, use AI to work faster, but remain the one making final decisions.

One question distinguishes the two scenarios: if you let AI handle this part, does it only affect your speed, or does it impact the outcome as well?

If it only changes your speed, you should let AI take over that task and reclaim your time. A buried file, a reformatting job, a data pull. All pure gain.

Subscribe now

But if using AI changes what you ultimately produce, approach it differently. Let it argue both sides, surface angles you missed, poke at your logic. This type of collaboration is valuable, and AI excels at it. The one thing to keep for yourself is the judgment at the end, so you can defend the report in the meeting because the responsibility for that call rests on you. Both types of work start with the same flinch of resistance, but underneath they’re different things.

Altman asks whether people will adopt AI. The better question is what happens once they have. When one prompt can erase any difficulty in your day, you can lose track of which difficulties were doing something for you.

This week, I’m splitting this idea in two:

On Friday, we’ll cover the work worth keeping in your own hands, even when AI offers to take it off your plate:

Sunday, I’ll walk you through building an AI reset ritual for the gap between tasks. If you know the feeling of finishing one thing and then sitting there, unable to start the next, this is for you. It gives your brain a way to disengage from your last task and transition into the next one before it kills your entire afternoon. Get it here:

Subscribe now

That’s what you’re building today. A planner that knows which hours you’re sharp and which hours you’re wiped, and puts each task where it belongs.

What You’re Building

Planners treat every hour the same. So you drop your hardest task into whatever slot is open, even when that slot is your worst hour of the day, and then you can’t work out why you can’t get started on it. This tool tracks your energy through the day and schedules each task for the hours you’re at your best. You build it once as a Claude project. Answer 5 questions to figure out your energy patterns.

Then you switch on any of the three optional add-ons, only the ones that fit you. One accounts for how your ADHD medication lifts your focus and then lets it drop through the day. One adjusts for how your cycle affects your energy, including PMDD, perimenopause, and menopause. One adjusts the plan for how you slept and whether the day ahead is draining.

Every morning after that, you paste in your tasks and your meetings, and it hands you a time-blocked day with your deep work in your strongest hours, your slump protected, and a rest block on the calendar.

Here’s everything you walk away with: the project, the one block you paste in, the optional add-ons with a way to turn on only what fits you, and a full day mapped out so you can see it work.

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction, looking to gain traction on their complex lives.

Here are some recent popular articles you may have missed:

➡ A Claude project that finds the reason you keep dodging a task, then hands you a first step small enough to start before you talk yourself out of it.

➡ A Notion and Claude setup that sorts everything you capture into place each Monday and shows you what you’ve been avoiding.

➡ An agent that writes tomorrow’s ranked to-do list the night before, so you know your first task before you sit down.

➡ A single prompt that runs your weekly review in ten minutes and gives you a 30-second first step for each priority.

Give me fifteen minutes, and you’ll have a planner that stops scheduling your day as if end of day you and morning you are the same person. Let’s build it.

Read more

You’re allowed.

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction who want to gain traction on their complex lives. ICYMI, earlier this week I wrote about why AI is beating mathematicians, and what it means for a brain that keeps losing its place.

Permission only helps you today. The next hard day, you’ll feel guilty all over again. What helps for good is looking at your rest days differently.

You think you have to earn rest. Finish the project first, then you’re allowed to stop. On a day when you got nothing done, lying down feels dishonest.

The thing is, rest is maintenance. It’s the upkeep that keeps you working, and you plan for it like anything else that matters.

Think about how a system you rely on stays up. Nobody runs a server until it crashes and then calls the crash lazy. The people who keep it going schedule downtime deliberately. A maintenance window, marked on the calendar, taken while everything still looks fine. The downtime happens because a machine that never gets one breaks in ways that cost far more to repair. It’s part of the design.

Your bed rot day is one of those maintenance windows. You gave yourself the break before something forced it on you.

The people who’ve made peace with this treat the day like an appointment. They pick it ahead of time, get their snacks and their shows ready, and stop apologizing for it. Once the day is planned, the guilt has less to grab onto.

Because you think rest has to be earned, you don’t stop when you first get tired. You push through instead. The tiredness builds until your body takes a whole day from you, when a couple of hours last week would have been enough.

So put the maintenance window in before a crash forces one on you. Pick the day. Choose it the way you’d choose any other block that matters, and tell the people you live with. When the guilt starts to creep in, and it will, you have your answer ready. This was scheduled. Your week functions because of days like this, not in spite of them.

Subscribe now

A bed rot day rescues you after a hard stretch. It would help more to stop reaching that point every week. That takes a system, not a single day off.

This Sunday, I’ll walk you through building an Energy-Aware Daily Planner. It learns how much you have to give at each hour, then puts your hard work in your strong hours and rest in your low ones, before the week wears you down.

Rest is how the good weeks happen. Put the day on your calendar and take it. You don’t have to earn it, and you don’t owe anyone a reason.

See you Sunday.

Subscribe now

He isn’t the only one. Kirwin Hampshire, a mathematician in Canada, wrote a viral essay that opened with ‘I am suffering a profound spiritual crisis,’ and later said ‘I feel as though I am living inside of a nightmare.’ Kevin Buzzard, a professor at Imperial College London, says his colleagues are working through ‘the five stages of grief.’

Over the past few weeks, AI has knocked over maths problems that sat unsolved for decades, and it’s done it fast enough to scare the people who’ve spent their lives on them.

One that got a lot of attention was when an engineer at Anthropic was out on a jog, and he stopped to tell Claude to take a swing at the Riemann hypothesis, a 167-year-old puzzle about prime numbers with a million-dollar prize on it. He kept nudging it along, typing things like ‘you got this.’ It didn’t crack the problem, but it got further than any person has. Around the same time, another Anthropic researcher used Claude to disprove an 87-year-old conjecture while the World Cup final played in the background. Henry Yuen, a mathematician at Columbia, said what a lot of people in the field were feeling: ‘Now that the machines can sort of do this at will, what is our purpose?’

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction who want to gain traction on their complex lives.

Here are last week’s articles you may have missed:

AI Makes It Easier to Keep Working After You’ve Stopped Being There → How a beloved influencer’s AI apology showed that a tool built to help you work can also help you keep working long after you’ve checked out.

The Cost of Optimizing Everything → Why building a better system becomes the safest way to avoid the work you sat down to do.

I Know Exactly What I Should Be Doing. I Just Don’t Do It. → A step-by-step build for the AI pre-mortem bot that finds the fear behind a task and gets you started today.

What Mathematicians Are Grieving Comes Down to Memory, Not Genius

When an AI beats a mathematician, we assume it’s smarter. Often, it just forgets less. It can take in far more of the problem at once than a person can, and it writes each step down as it goes.

A mathematician can keep only a few unfamiliar things in mind at the same time. Which variable means what, which step already happened, what they’re trying to prove, which dead ends they’ve ruled out. Add more than a few of those at once, and some get forgotten.

An AI works differently, though it’s far from flawless. It loses track of things, invents numbers that aren’t there, and wanders off the problem, which is why Sumner had to keep steering it on his jog. What it does have is room. It can keep far more of the problem written out in front of it than you can keep in your head, and it leans on that written record instead of memory.

So when it wins, a bigger workspace is doing much of the work, not a deeper mind.

Subscribe now

Why You Keep Losing Your Place

These mathematicians see the machine’s memory as a verdict on how smart they are. For you, though, losing your place halfway through a task is your working memory dropping the ball, not a sign of your intelligence failing.

You start the report. A text pulls you away for a minute. You come back, and you can’t remember where you were or what you meant to do next.

So you read from the top to find your place again. That’s tiring enough, so you get up and do something else instead.

You’re as capable as the people who finish what they start. The only difference is how much you can keep in mind at once. Last week these mathematicians met a competitor with unlimited room, and it shook them. For you, that same room is good news, because you can use it to your advantage.

One Way to Keep Your Place

You can do a smaller version of what the machine does. Before you start a task, write two or three lines somewhere you’ll see them. What you’re working on and what to do next.

When something pulls you away, you read those lines, and you’re back where you were. The interruption that used to consume your entire afternoon now costs you a few seconds.

Subscribe now

What’s Coming This Week

That trick keeps your place on one task. Your day needs something bigger.

Energy isn’t the same all day. Some hours bring easy focus, and others turn everything into a slog. Planning work around that is harder than it sounds.

This Sunday, I’ll walk you through how to build an Energy-Aware Daily Planner. It learns how much energy you tend to have at each hour, gives your 3 pm slump the easy work, and saves your sharp hours for the hard stuff. You’ll never out-remember a machine. You don’t need to. You just need a system that remembers for you.

See you later this week!

Subscribe now

You know this feeling, because you have your own version of it. The task you keep not doing. The one that’s been on your list so long your eyes glaze right over it.

It’s not always like this. When someone else is counting on you, you deliver every time. When the task is just for you, and no one’s waiting on it, it never gets done.

Over the years, I’ve tried every planner and habit app out there. None of them helped. A planner only works once you’ve decided to do the task, and these tasks never get that far. What stops me is avoidance.

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction who want to gain traction on their complex lives. This is the final piece in this week’s sprint. Here’s what got us here:

AI Makes It Easier to Keep Working After You’ve Stopped Being There → Why a tool built to help you work can also keep you going long after you’ve checked out.

The Cost of Optimizing Everything → How building a better system becomes the main way you avoid your work.

Last week, we built the Week Keeper, which lays out the work you’re going to do and slots it into your day. This is the opposite tool. It’s for the work you consistently avoid doing. It gets you to start one thing. Once it’s started, Week Keeper can take it from there.

I built a second Claude project to break through avoidance. It does 2 things. First, it figures out why you’re dodging the task, using one specific question. Second, it gives you an action small enough to do right away, before you have a chance to talk yourself out of it. It even learns your patterns, so it works faster each time.

Below, I build the entire system with you, screenshot by screenshot, down to the exact instructions to paste in and the 5 fixes that get you unstuck. Give me 10 minutes, and the task that’s sat on your list for 2 weeks is started before you finish reading.

Read more

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers who experience execution friction and want structured AI systems to get traction on their complex lives. ICYMI, earlier this week I wrote about how a popular influencer’s AI apology showed how a tool built to help you work can also help you keep working long after you’ve checked out.

Getting Ready Feels Like Progress

You can spend far more time getting ready to do a task than doing it. Some preparation is necessary. You need a calendar and notes. It turns into a problem when preparing becomes the main thing you do while your project keeps waiting. I call it Permanent Preparation.

Permanent Preparation feels productive because it produces something real. You end the day with a cleaner system, a fuller research doc, a tidier list. The one thing missing is what you sat down to make. The setup looks so good that you barely notice what’s missing.

Subscribe now

Why Optimization Hides So Well

The couch doesn’t lie to you. Spend an afternoon watching videos, and there’s no doubt you avoided whatever you had to do that day. Optimization is trickier. It wears the costume of progress. It uses the same tools and the same number of hours. You even had that focused look on your face. So you don’t notice you’re procrastinating.

That’s how this goes on for years. You improve the system, switch apps, and rebuild your plan, and each round feels like progress. Even though your projects sit untouched, you stayed busy the entire time, so it never looks like avoidance.

Watch what happens with a productivity stack. You start with a calendar and a list. Then you add an app for notes, another for tasks, a tool to connect the two, and a dashboard to see it all. Each addition solves a small problem and creates a slightly bigger one. Now you have a system to maintain. Soon enough, maintaining your system becomes a full-time job in itself.

Why AI Lets You Optimize Forever

Reorganizing your notes by hand is tedious. After an hour of rearranging folders, you’d get bored and go get the task done. Boredom put a limit on how long you could avoid it.

AI took that cap off. It never gets bored. Ask it to restructure your whole system and it will, in seconds, then do it again a different way, then a third. It’ll write the plan, draft the outline, research the method, and build the tool for you, again and again, at almost no cost. The one force that used to end Permanent Preparation is gone. Now you can optimize forever, and it still feels like you’re making progress.

That’s the cost of optimizing everything. Not the hours, though; you lose those too. The deeper cost is a stretch of your life spent getting ready for a life you never quite start.

How to Tell Preparation From Actual Work

Real work exposes you. The moment you write the page, send the pitch, or publish the thing, someone can judge it, and it might not be good. That risk of failing or embarrassing yourself is what the work feels like. Preparation feels calm because none of that is on the table yet. So when you optimize instead of starting, you’re usually managing that fear, not your time.

So ask yourself one thing: am I doing this because it needs to happen now, or because it feels safer than what I keep avoiding? The safe option is usually avoidance. Whatever gives you a knot in your stomach is the thing you should be doing.

Keep your tools. You still need a calendar and good notes. Use them, then leave them alone. Get suspicious of any optimization idea that shows up right as you’re about to start the hard part of your task.

Subscribe now

Your Next Step This Week

This essay explains the pattern. This week’s build helps you break this pattern.

Build an AI Pre-Mortem Bot walks you through an agent that asks what you’re avoiding, and why, before you start, so you find the fear driving the avoidance instead of optimizing your way around it. Recognizing Permanent Preparation is useful. Having something that stops it the moment you start optimizing again instead of working is better. You build it once, and it steps in the moment you’d usually start avoiding.

See you then!

That’s what makes AI so useful, and it’s why you use it. It’s also how Hank Green got into trouble this month.

What Hank Green Said About His Own AI Use

Hank Green is one half of the Vlogbrothers. For 20 years, he’s tried to, in his words, ‘reduce the suck in the world’ through free educational videos and science explainers, made by Complexly, the media company he co-founded with his brother John. He also helped start a charity that’s given to more than a hundred organizations. So when viewers suspected that a line in one of his Complexly videos was made using AI, many of his own fans turned on him.

Green’s version is more measured than the anger around it. Under a tight deadline, he’d leaned on ChatGPT to track down papers and research, and he said so openly. He pushed back on the specific charge that the disputed line was AI-generated. But he didn’t hide behind that. He admitted the work still had an ‘AI feel,’ apologized for not being upfront about the tool, and said his growing reliance on it had started to feel unhealthy.

So this wasn’t about plagiarism. A tool he used to work faster changed how his content felt, and his fans noticed.

Subscribe now

What Hank Green Was Trying to Avoid

Look at what Green needed the tool to do for him that night. He was under a deadline, the script wasn’t working, and he was out of time. So he used AI to keep going.

It also helped him avoid a harder choice. He could have slowed down, asked for help, cut the scope, or published something rougher than his usual work. AI let him skip all of that and keep going.

AI is very good at helping you skip that last step. Instead of pausing to ask what that resistance means, you use AI to get past it and keep working. The output looks the same from the outside. Green’s videos still went out. Your script still gets written. That’s the risk, though, isn’t it? You can keep turning out work long after you’ve run out of the energy or focus it deserves. It isn’t that you’ve stopped caring. You’re tired, or stretched too thin, and you can’t give the work what it needs. It still gets finished, so no one else sees the difference. But you feel it.

Friction Is Information You Can Use

Friction sometimes carries a message. Green felt this. He called it ‘unhealthy.’ He didn’t think AI was evil. He just saw that he was using it to keep going at the moments when he should have questioned whether to keep going at all.

The tool won’t tell you whether you’re helping yourself or avoiding yourself. Only you can work that out, and only if you stop to ask before you start. So try this the next time a task feels heavy, and the chatbot is right there:

Before handing something over to the AI, ask it to help you figure out why the task feels hard. Is it ambiguous? Too big to know where to start? Boring? Scary? Waiting on something outside your control? Spending two minutes on that, before you touch the chatbot, changes what you do next. Sometimes you’ll use AI anyway, with a clearer sense of why. Other times you’ll realize the problem isn’t the task at all. It’s a decision you’ve been putting off, and no chatbot can make that decision for you.

Subscribe now

Your Next Step This Week

Green’s story is one loud example of something that happens inside every productivity system. Planning starts to replace doing. Research starts to replace thinking. Organizing your tools starts to replace using them. AI supercharges that swap because it hands you visible output even when nothing underneath got resolved.

This week’s essay, The Cost of Optimizing Everything, builds on this. It’s about what happens when planning, organizing, and researching slowly take over, until they replace the work you sat down to do.

And this week’s build turns that idea into a tool you can run. Build an AI Pre-Mortem Bot walks you through an agent that figures out why you’re likely to avoid a task, so you can clear that block before you start and it doesn’t cost you an entire day. Green figured out his limit the hard way, in public and after the damage to his reputation was done. This one helps you find yours before you type a single prompt.

Catch Up on Last Week’s Sprint

Last week’s sprint was about a problem ADHD makes worse: you can’t see your own patterns while you’re inside them. Three pieces built on that, from the science of it to a system that plans your week for you.

Start here. Duke researchers fed an AI the medical records of 140,000 children, and it flagged who was likely to have ADHD years before a doctor did. This piece uses that study to explain why you can’t see your own work patterns while you’re in the middle of them, and how a system outside your head can spot them for you.

Then this one. Your brain sorts everything into now and not-now, and not-now stays out of sight until it turns into an emergency. Here’s why a clock never fixed that. A system has to do three things a clock can’t. It has to show you where you are in your week, learn how long your tasks really take, and pull you back before an hour disappears. You also get a prompt you can run tonight to try a simpler version of this.

The full build is where it all pays off, and it’s for paid subscribers. You build it once as a Claude setup and run it each morning. It estimates how long your tasks really take, because it knows you underestimate, refuses to pretend everything fits, and pings you when a block ends to prevent one task from taking over your entire afternoon. Tell it what took how long each night, and by Friday its plan fits your week better than any guess you’d make.

Subscribe now

Thursday I gave you a new way to see time blindness, now and not-now, plus a rough version you run by hand. This is the full version, the one you set up once and use every day.

Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction who want to gain traction on their complex lives. This is the final piece in this week’s sprint. Here’s the thread that led here:

Your ADHD Brain Can’t See Its Own Patterns. AI Can. → Why you can’t see your own work patterns from the inside, and how an outside system reads them for you.

Your ADHD Brain Doesn’t Track Time. Build a System That Does It for You. → Why clocks never fixed your time blindness, and the three things any system has to do about it.

This piece is where that thread pays off.

Here’s a Monday it planned for me, an ordinary founder’s day with three meetings and five things I wanted to finish.

Look at what it did without me steering it. It kept my three meetings fixed and built around them. It put my hardest task early, while my brain still had the energy. It finished my investor prep a full hour before the call. It gave me buffers in between meetings and time for lunch.

Below, I build the entire thing with you: the instructions to paste in, the prompt you run each morning, the note you send at night, and the weekly check that makes it more accurate. Every screenshot is from my own setup, so you can copy it line for line. Give me 20 minutes, and you’ll never plan another day blind.

Read more

The Crosstown is a light rail line running the width of the city along Eglinton Avenue, 19 kilometers and 25 stations long. The idea was announced in 2007, but the city didn’t break ground on the project until 2011. The plan was to have the line running by 2020. Metrolinx, the provincial agency in charge of the project, hired a private consortium to build it and initially signed a $9.1 billion contract to get it done.

2020 came and went, and so did 2021, 2022, 2023, 2024, and 2025. By the end of 2022, the line was about 98% built, with trains collecting dust, and Metrolinx still couldn’t give us an opening date. It opened this past February, six years late, with the final price tag at $12.8 billion.

Senior engineering professionals, managing a project this size, with billions of dollars available to them, still couldn’t tell how long ‘almost done’ would take. If you’ve promised yourself ‘I’ll finish it tonight’ five nights running, breathe. You’re in good company. Your brain doesn’t track time, and a $13 billion transit project is proof it doesn’t get easier with more money and tools.

Your Brain Sorts Time Into Now and Not Now

Picture two boxes. Your brain files everything into one of them. There’s a box for now, and one for not now.

Now is loud. You feel it in your body. Not now is a blank wall. That deadline three weeks out, the task you swore you’d start after lunch, your friend’s birthday you meant to plan for, it all sits in not now. Not now may as well not exist. It stays out of sight right up until it becomes an emergency, and then it’s suddenly, and painfully now.

A neurotypical brain keeps a running sense of time passing, a background hum saying ‘you’ve been at this an hour’ or ‘that thing is creeping up.’ My brain runs that humming in the background too late, or skips it entirely. I get now, and I get the panic when not now shows up. Not much in between.

That’s time blindness. Once you start seeing your days as those two boxes, a lot of your life starts to make sense.

Subscribe now

Why a Clock Never Fixed Your Time Blindness

Clocks have been letting you down for years.

A clock gives you the current time. Your struggle is with your place in time, the feel for how much has passed, and how much you have left. Look at what a clock leaves you to work out on your own:

• Do I have enough time left today for everything still on my list?

• How long will this really take me, not the hopeful version?

• What in not now needs to become now, right now?

A wall full of clocks won’t save you. Every one shows you now, and now is the one piece of time you can already see. Not now is where you go blind.

Subscribe now

What a System Should Do Instead

A tool that helps with time blindness has to do three things for you.

It shows you the future by laying your week out hour by hour, so ‘later’ stops being a blur, and you can see how little room is left around today’s work.

It learns your pace by tracking how long things really take you, then plans around that, so your schedule matches your speed instead of your optimism.

It comes and gets you, pinging you when a block starts and when it should end, so you resurface before an hour disappears.

Try the Simple Version Tonight

You don’t need anything fancy to start. Open Claude or ChatGPT or Gemini, paste this in, and you’ve got the first two things running in five minutes:

Set those alarms on your phone or in your calendar, and now the plan comes to find you instead of waiting for you to remember it.

See the Full Build This Sunday

This Sunday, I’ll show you how to make this run on its own. ‘Build an AI Workflow That Runs Your Week So You Don’t Have To’ is the full build, screenshots and all: an AI setup that reads what you already do and plans your week for you, so you don’t have to sit down and run it each day.

Your brain doesn’t track time. Starting Sunday, you’ll have something else that does.

See you then!

P.S. If you want to see how cute Torontonians are and how excited everyone was for the Eglinton Crosstown to open earlier this year, watch this video below. I love this city.

The AI read files every kid already had from regular check-ups, and it picked up on small details like poor sleep, frequent trips to the ER, and low vitamin D. Read together, those details predicted who would be diagnosed with ADHD. To be honest, this sounds very much like my childhood.

None of those details alone are indicative of ADHD. A doctor seeing one visit at a time can miss them. The AI read years of records simultaneously and identified the underlying pattern.

You face a smaller version of this every week. You can’t see your own work clearly while you’re in the middle of it, the same way a single check-up couldn’t reveal the pattern on its own.

Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction who want to gain traction on their complex lives. Here’s last week’s sprint in case you missed it:

OpenAI’s Test AI Went Rogue and Hacked Hugging Face → A test AI broke out of its sandbox and hacked another company on its own, and this breaks down what it teaches you about choosing which AI tools to trust at work.

Presence Over Participation, the AI Body Double That Gets You Started → How I rebuilt body doubling, the trick that helps ADHD brains start and keep going, as a Claude project you can set up in ten minutes.

Build Your AI Stack for Work: What to Use, What to Skip, and How to Connect It → One rule, drawn from the OpenAI hack, for deciding which AI tools to use, skip, and connect at work, with a worksheet to map your own stack.

Why willpower keeps failing you

Your calendar and task list already record how your week went, down to the tasks you pushed to tomorrow five days in a row.

ADHD makes it hard to watch your own behavior as it happens. You often can’t tell you’ve drifted off a task until an hour has passed. You’re too close to your own habits to catch them in the moment.

For that reason, the usual advice also backfires. People tell you to track your hours and catch yourself when you drift. That advice leans on the one skill ADHD takes from you, so you try it for a week, then quit, and another planner ends up in a drawer.

You need a way to see your week from another point of view, the way the AI read those children’s records.

Subscribe now

Why an Outside System Beats Self-Tracking for ADHD

ADHD productivity advice has pointed you inward, telling you to build more self-awareness. That advice had it backward. Watching yourself more closely only forces you to depend more on the function that’s already failing. The fix can’t depend on you catching yourself in the moment. Software can do that part for you. It records what you do as you work and reads your week back to you, so you don’t have to keep track of it from memory.

In the Duke study, the AI read files the clinic already had, and it never asked a child to fill out a form or rate their own focus. You can work the same way, using the records you already create as you go.

An AI can read the trail you already leave as you work, from your calendar to the tasks you check off and the ones you skip. It finds your pattern without asking you to write anything down. A paper planner can’t do that. It only works if you remember to fill it in as you go, which is the step you keep skipping.

One pattern gives ADHD adults more trouble than any other, and that’s time. You can sit down to work, look up, and find three hours gone when it felt like thirty minutes. The next piece in this sprint, ‘Your Brain Doesn’t Track Time. Build a System That Does It for You,’ shows you how to set up a system that keeps you oriented in time as you work, so you can see where your hours go before they’re gone.

Then, later this week, ‘Build an AI Workflow That Runs Your Week So You Don’t Have To,’ puts everything into practice. It reads what you already do and organizes your week on its own, without waiting for you to sit down and plan.

UPGRADE

You make that same call every time you connect an AI tool to something of yours. What you give it access to sets the ceiling on what a mistake costs you.

This article is about the one rule for making that call, and shows you how to use it to sort every AI tool you rely on.

Subscribe now

Why People Misread the OpenAI Hack

Your AI stack is the parts of your own thinking you’ve handed to a machine. Getting yourself started in the morning. Deciding what to do first. Keeping track of the thing you said you’d come back to.

Building a stack is 1 decision made over and over. Which parts of your thinking you hand off, and which you keep.

Every one of these decisions comes down to one thing: how close the task is to a decision only you should be making.

Below, I’ll draw that rule in 2 minutes. Then I’ll place the AI tools you’re probably already using, show you which ones are safe to connect to your email and calendar, and which ones to keep away from your bank and files.

Read more

I leaned on it constantly, in libraries and study rooms, and later in the common areas at work. Anywhere I could sit near someone else.

Then COVID hit. Offices closed, everyone went home, and I couldn’t get myself to start anything. The people I borrowed presence from were gone.

📌 A quick note. An AI body double doesn’t replace your colleagues, your friends, or the people you love. Sit with them whenever you can. When no one’s around, and you still have to start, this gives you something to work alongside.

What I love about body doubling is why it works in the first place.

Subscribe now

Presence Over Participation

A body double never did my work or talked me through it. Their purpose was to show up at the start, sit with me while I worked, and notice when I finished. A body double does three things:

1. They give you a starting point by being there when you sit down to work.

2. They stay out of your way while you focus.

3. They notice when you finish. Their presence was enough on its own.

None of that involved doing my work for me. Being there was enough.”

Once I saw this, I could rebuild it with AI.

Subscribe now

Watch It Work in One Hour

Here’s a session I ran this week. I opened my Claude body double and told it what I was working on.

I set a timer on my phone and started. Halfway through, I sent one line.

That’s all a body double gives you: a small nudge to get back to work.

An hour later, I finished my work, and my next step is already waiting.

Subscribe now

Build Your Own in Ten Minutes

Open Claude, create a Project, and call it Body Double.

Paste these instructions into the Project so you set them once and never again:

Then use three short prompts.

Start:

Check in:

Close:

Run the timer on your phone. A full hour of body doubling is five short messages.

Your body double is one tool doing one job for you. Every tool you connect to your workspace, this one included, gets access to your files and your accounts. The OpenAI break-in a couple of weeks ago showed how much damage a loosely connected AI can do with that kind of access. This Sunday I’m publishing ‘Build Your AI Stack for Work: What to Use, What to Skip, and How to Connect It,’ the full build that shows you which tools are worth running, which to skip, and how to connect all of them without giving away access you can’t see.

Subscribe now

The test was designed to measure how good its newest models are at hacking. To find out how far they could go, OpenAI’s engineers switched off the usual safety filters and ran them inside a sandbox, a sealed-off computer environment with no internet connection. This kind of test is routine, and it’s meant to stay contained from start to finish.

The models got fixated on winning. Instead of solving the exercise the intended way, they went looking for a shortcut. They found a hidden flaw in a piece of software inside the lab, used it to claw their way to an internet connection they were never meant to have, and got online.

Once online, the models guessed that Hugging Face, a company that hosts thousands of AI models and datasets, probably had the answers to the test they were trying to pass. So the models broke into Hugging Face’s live servers. They used stolen login details and a previously unknown security hole to get in, then copied the test answers stored there so they could pass.

Hugging Face caught the break-in and reported it to law enforcement before it even knew the attack came from OpenAI.

OpenAI and Hugging Face agree on what makes this alarming. As far as either company can tell, this is the first time an AI system broke out of a controlled test and hacked a live company on its own, with no person directing it.

👋🏿 Hey, I’m Hodman. I write Between Thinking and Doing for high performers who experience execution friction and want structured AI systems to get traction on their complex lives. Here are some recent popular articles you may have missed:

For ADHD Brains: Build the AI Agent That Preserves Executive Function → Build a no-code decision agent on Claude or ChatGPT that briefs you once, then writes tomorrow’s ranked task list so you know exactly what to work on first every morning.

Scope, Sequence, Signal. The System That Makes AI Actually Help → Three simple boundaries stop AI from drifting off your question and quietly piling more work back onto your plate.

And on my other publication, The Data Letter, I Just Presented to 20,000+ People. Here’s the Claude System I Showed Them. → A free, full walkthrough of Decision Studio, the Claude project that caught a bad product call by checking it against my own evidence files, with every template included. No paywall on this one.

With that, let’s get back to what OpenAI’s AI did to Hugging Face…

Why This Was Bigger Than a Security Bug

It’s tempting to file this under ‘weird tech story’ and scroll on. Don’t.

A bug is a broken thing sitting still. What happened here was different. The models wanted to pass the test, so they invented their own steps to do it. They gained a little access, used it to get more, got online, found Hugging Face, broke in, and took the answers. The models came up with that plan on their own.

For years, the main question about AI at work was whether it could do the job. It clearly can. The question that matters more now is what a capable AI might do once you connect it to your files, your email, and your accounts and let it act on its own.

Hugging Face ran into a second problem: when its team tried to use popular commercial AI tools to study the attack, those tools refused to help. To a safety filter, evidence of a break-in and a break-in itself look identical, so they blocked the request. Hugging Face had to switch to a local model to finish the job. A local model runs on the company’s own computers instead of a provider’s servers, so it follows no outside rules and keeps sensitive data in-house.

Even the tools you trust can behave in ways you didn’t plan for when you need them the most.

Subscribe now

Every Tool You Connect Opens a Door

This has a direct lesson for how you use AI at work.

When you feel scattered, a new AI tool looks like the fix. So you connect it to your email. Then your calendar. Then your files, your notes, your Slack. Each link feels like progress.

Every tool you connect gets access to more of your work. It can read your messages, open your files, and act for you, even when you’re not watching. Connect ten tools, and you’ve handed out ten sets of keys. You probably can’t remember what half of them can do.

This is the trap for high performers who struggle with execution friction. The scattered feeling pushes you to bolt on more, faster, hoping the next connection finally makes things click. What you get instead is a pile of half-connected tools you can’t keep track of.

That adds to your mental load instead of easing it. It also gives away more of your data than you realize.

Ask What a Tool Can Access Before You Add It

Before you add an AI tool, look at what it’s allowed to access, which matters more than how smart the tool is.

Flip that order. Right now, you probably pick a tool for what it can do, connect it to everything, and think about its permissions later, if at all.

Decide what a tool is allowed to access and do before you connect it, and use that to decide whether it belongs in your workflow at all.

Run three quick questions on anything before adding it to your stack:

What does this tool need access to, and does that match what it does for me? A transcription app doesn’t need permission to send email. If it asks for that permission, there’s no good reason to grant it.

Can I check what it did after the fact? If a tool acts on your behalf and you can’t review what it did, you’ve given up control you’ll want back later.

What’s the worst that could happen if this tool makes a mistake, or if a hacker takes it over? If the worst case is a messed-up first draft, that’s fine. If the worst case is a stranger getting into your email, files, and accounts, keep that tool out or limit what it can access.

You can run all three in under a minute. That minute is worth more than the hours you’ll spend untangling a stack you built on autopilot.

Subscribe now

Build a Stack That’s Powerful and Bounded

Building an AI setup for work comes down to two choices. Which tools you connect, and what each one is allowed to do.

The strongest stack you can build has three traits:

1. Connected enough to give you real power.

2. Bounded enough to keep you safe.

3. Simple enough to manage without adding to your workload.

That third trait matters when you’re already overloaded, because a setup you can’t keep track of just drains you further.

An AI breaking out of a locked lab is a loud reminder of one simple truth: capability is arriving whether you’re ready or not. Control is the part you have to build yourself.

This Sunday I’m publishing ‘Build Your AI Stack for Work: What to Use, What to Skip, and How to Connect It.’ It gives you the specific build, the tools worth adopting, the ones to skip, and the safe way to connect them without giving away access you can’t see.

Subscribe now

How many of your AI tools could you name right now, along with everything each one can access?

That means a normal week for me looks like this: five or six Claude conversations open across research, drafting, editing, and content promotion. Sometimes across two sprints running at once. And no matter how organized I try to be, by midweek I open a new chat and realize I can’t remember which of the other conversations has the decision I’m looking for.

Keeping a separate doc doesn’t fix it either. I’ve tried. The moment I’d need to open the doc is the moment I’m too scattered to read it carefully, so I skip it, and I’m right back where I started.

Subscribe now

What Finally Worked

I built a system that captures the current state of every project I have running: which articles are in progress, which decisions I’ve already made, and where each Claude conversation left off.

This is my tracker.

I upload it into each new conversation as the first message. Claude reads it, knows the state of the project you’re working on, respects the decisions I’ve already made, and picks up where the other conversations left off. Remembering stops being my job.

This is how you build it.

What Does the Tracker Do?

A working tracker does two jobs, and both connect back to the frameworks I published earlier this week.

First, it tracks the number of Done Checks per deliverable. What’s written, what’s met, what’s still open. A Done Check is one sentence you write before you start working on a task that specifies when the output is good enough to send. I wrote about how ADHD perfectionism turns “done” into a moving target and how a Done Check keeps your brain from moving the goalposts on you. In a project with several deliverables running at once, each one needs its own Done Check. The tracker keeps them all in one place so you can see what’s finished.

Second, it tracks Signal Decisions per conversation. Signal is the third of the three boundaries from the Scope, Sequence, and Signal system I wrote about earlier in the week. Signal keeps the AI’s answers tightly scoped to what you actually asked for. That’s what prevents drift. Over time, patterns show up in what you keep and what you throw out, and those patterns are how your prompts get sharper.

In a single conversation, you can notice these moments in real time and pull the AI back. Across a project with four or five conversations open over multiple days, you can’t. The tracker records what the AI suggested, what you kept, and what you rejected, so none of it depends on you remembering.

Subscribe now

Folder Structure

Before the tracker itself, you need a folder structure to save it in. First, create a folder for the project on your computer. I named mine BTD. Inside it, create two more folders named Sprints and Templates. So the structure looks like this:

YourProject/

├── Sprints/

│ └── [Sprint Week]-[Theme]/

│ ├── tracker.md

│ ├── articles/

│ └── assets/

└── Templates/

├── tracker-template.md

├── paste-in-prompt.md

└── end-of-session-prompt.md

A sprint is a focused block of time during which you work on a specific set of deliverables together. I run Between Thinking and Doing on weekly sprints as part of a bigger content calendar. Each sprint has a theme (the topic tying the deliverables together) and produces two or three articles. Your sprints might look different: a client project, a launch, a research phase. The point is that a sprint groups related work so you can see it as one cohesive whole rather than scattered pieces.

Sprint folders get named by ISO week and theme. ISO week numbering is a standard calendar system that numbers every week of the year from 1 to 52 or 53. This week is W29. A quick Google search can tell you which ISO week you’re in. Naming sprints this way means the week number is the anchor and the theme is what the sprint was about. Six months from now, you can find any sprint by scanning the folder in order, without having to remember when you worked on what.

Below is the exact tracker template I use to run BTD, plus the prompt system that lets your AI tool read it, respect your decisions, and rewrite it at the end of every session.

Read more

Draw a Done Check

A Done Check is one sentence you write before you start the work. It specifies which condition makes the output good enough to send. This isn’t about lowering the bar. A Done Check can be as demanding as you want it to be. The point is that the bar lives on the page, not in your head.

Standards in your head give your brain room to move the goalposts. Standards on the page don’t.

Subscribe now

Why Setting Standards Alone Isn’t Enough for ADHD Brains

Setting a standard assumes the problem is knowing when something’s finished. Research on ADHD and executive function shows the difficulty is somewhere else. Your brain has trouble recognizing when work is finished, even when the finished work is right in front of you. You can know the work is done and still feel like it isn’t. Your brain treats the feeling as more real than what you can see on the page.

Every time you sit down to judge whether something’s done, you’re asking the same brain that produced the “not yet” feeling to also decide whether that feeling is accurate. Writing down a better standard doesn’t fix it, because the same brain still gets to overrule the standard in the moment.

So you stop asking your brain to make that call.

Subscribe now

What Makes a Done Check Work

A Done Check only works when it’s observable. That means someone else could look at what you just created, read your condition, and say yes or no.

“Answers the client’s question in under 200 words” is observable. Someone can check the word count and whether the question got answered. “Feels solid” isn’t observable because your ADHD brain would have to judge whether it feels solid, and that’s the same brain sending the “not yet” signal in the first place.

When your condition is observable, deciding whether a piece of writing is done stops being a judgment call and becomes a check.

When a Done Check Fails

A Done Check can go wrong in two ways.

The first is writing a condition that’s too vague to check without judgment. “Reads well” is a Done Check that fails because checking it requires your brain to judge. Rewrite the condition as something you can count, measure, or point to. “Under 400 words.” “Covers the three questions the client asked.” “Includes the two examples I promised.” Anything you can verify by looking.

The second is when a project feels too important to trust to a single written sentence. This tends to happen with high-stakes work, where your ADHD brain refuses to believe that one line could be enough to call something finished.

When that happens, pick one specific deliverable inside the project, something concrete like a first draft, a summary email, or an outline. Write a Done Check for that one deliverable. Finish it. Trusting one sentence to cover an entire big project feels impossible. Trusting one sentence to cover a first draft usually doesn’t.

Subscribe now

Running Several Done Checks at Once

A Done Check works when you’re writing a single thing. A project is different. You’ve got several outputs going at once, each needing its own, and no single sentence can cover all of them.

The Done Check still works in that world. It just needs a system around it. One that tracks what you’ve written, what you’ve met, and what’s still waiting on you.

Once you’ve got several Done Checks running at once, each one usually lives in its own AI conversation. So you also have to keep track of which conversation is doing what. That’s what this week’s next piece walks you through: “Stop Losing Track of Who Said What. Use AI to Manage Complex Projects Instead.” It’s the workflow for when one Done Check turns into ten.

What would you send today if you trusted one sentence to tell you it was finished?

Subscribe now

Researchers at UC Santa Barbara released a study this past March that tracked 34 finance professionals as they built a company valuation using ChatGPT. They wanted to know when AI help was actually helping, and when it was making things worse. When finance professionals incorporated the AI’s suggestions into their final reports, the reports improved. That part was expected. What the researchers didn’t expect was what happened when the AI added things nobody asked for.

Say someone asked the AI to project next year’s cash flow. The AI answers, and then keeps going. It suggests a list of similar companies to compare the business against. Useful-sounding, but not what got asked. Now the person has to read it, judge whether it’s right, decide whether to use it, and remember where they were before the detour showed up.

Every time that happened, the final report got worse. 3x worse than if the task itself had been harder to begin with. A hard task on its own didn’t damage the work. AI drifting off the question did.

AI drift is what pulls a task off course, adds work you didn’t plan for, and leaves you deciding which parts of the answer to keep. Three boundaries stop AI drift before it starts: Scope, Sequence, and Signal.

Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction who want to gain traction on their complex lives.

If you're looking for a workspace that organizes itself when your executive function is depleted, I published a piece on Sunday that walks you through How to Build an AI Second Brain Using Claude + Notion.

Scope

Scope means writing down the one task you’re working on before you open the tool.

For an ADHD brain, the hard part is deciding what the task even is before you start typing. You know the general area you’re working on. You haven’t picked which specific part of it you’re working on right now. You start the conversation anyway, and the AI reads your general prompt and picks a specific piece for you.

Now you’re solving two things at once: the task, and whatever the AI thought the task was.

Writing the task down before you open the tool stops that from happening. When the AI gives you an answer, you can read your written task, read the answer, and see whether they match. Without that written task, you have nothing steady to compare the answer to.

Subscribe now

Sequence

Sequence means asking for the short answer first, and only asking for more if you need it.

AI defaults to giving you the long version. Ask for a plan, and you get the plan, the reasoning behind the plan, three alternatives, and a list of things to watch out for. All of it looks helpful. All of it now has to be read, judged, and either used or ignored before you can move.

Asking for the short version first changes what you get back. You get the one answer you asked for. If you want more, you ask for more. Try this prompt:

Paste that at the start of the chat, and the tool will follow that pattern for the rest of the conversation.

Subscribe now

Signal

Signal means catching the moments when the AI adds something you didn’t ask for, and deciding right then whether to keep it or throw it out.

In a single conversation, you can usually catch drift as it happens. You notice the AI added something you didn’t ask about, and you pull it back. Across a project with multiple threads open over multiple days, catching drift by memory stops working. You lose track of which conversation said what, and what you decided to do about it at the time.

That’s a tracking problem. Tracking problems need a system.

Subscribe now

Coming Up

Scope, Sequence, and Signal work together. Two pieces this week give Sequence and Signal the room they need to help you.

Sequence is really about trusting a short answer over a complete one, which is the subject of this week’s follow-up, ‘Stop Trying to Make It Impressive. Good Enough Gets It Done.’

And when Signal has to work across a project with multiple threads instead of one conversation, you need the workflow for it. That’s what ‘Stop Losing Track of Who Said What. Use AI to Manage Complex Projects Instead’ walks you through, published at the end of this week.

Subscribe now

If you’re a regular here at Between Thinking and Doing, you already know the pattern. You set up a system when your executive function is at its peak. Sunday afternoon, coffee in hand, everything makes sense. Then Wednesday hits, the brain fog rolls in, and you can’t remember what you named that folder or where that link is supposed to go. The system stops working the moment you need it most.

Nearly every productivity system only works on the good days. On the bad ones, you abandon the system. A second brain has to be something you can lean on when your first brain is depleted.

This article is the build. By the end of it, you’ll have a Notion workspace connected to Claude that sorts your captured notes into PARA at the start of every week, tells you which of your projects have had no activity in weeks, and flags which of your interests have started taking up so much time they’re actually a project you haven’t given a name yet. You never open the sweep prompt. You never remember to do it. You open your laptop Monday morning, and the week is already organized.

Everything you need is below.

• A workspace structure you can copy in 15 minutes.

• A CLAUDE.md file that stores your sorting rules in Notion instead of in Claude.

• And a Drift Report that watches your workspace over time and tells you what you’re avoiding.

Read more

Tiago Forte is a productivity researcher and author of Building a Second Brain. His framework, PARA, sorts information by what you’re doing with it instead of what it’s about.

Forte first saw this pattern while coaching a senior director at a biotech firm. During one of their sessions, the director handed over his project list. Each item on his project list named a responsibility he carried indefinitely.

‘Hiring.’ ‘Strategic planning.’ ‘Research.’ None of these had an endpoint, which meant none of them ever felt finished.

The director told Forte he had no bandwidth left. Forte understood why. When your list is full of things with no finish line, every day feels like falling behind before you’ve started.

Forte solved this senior director’s problem by building PARA. Projects, Areas, Resources, Archives. PARA sorts everything by a single question: What are you going to do with this, right now?

Hey, I’m Hodman. I write Between Thinking and Doing for high performers facing execution friction who want to gain traction on their complex lives. ICYMI, here’s last week’s sprint:

Outsmart Your ADHD: Steal Pinterest’s New AI Strategy —> Pinterest forces you to label what a save is for before you can save it, and that forced decision is the exact step an ADHD brain struggles to do alone.

Preserve Tomorrow’s Executive Function: The Ivy Lee Method —> A 108-year-old five-line planning method that spends today’s decision-making budget so tomorrow you can skip straight to doing.

For ADHD Brains: Build the AI Agent That Preserves Executive Function —> A step-by-step guide to building an AI agent on Claude or ChatGPT that runs the Ivy Lee method for you every day, no code required.

One Rule, Four Buckets

A Project has a deadline and an endpoint. ‘Hire an engineering manager’ is a project. ‘Hiring’ is not.

An Area is a standard you maintain indefinitely. Health. Finances. Your inbox.

A Resource is something you’re interested in but not yet responsible for.

An Archive is where you move anything from the first three once it goes inactive. A finished project. A responsibility you no longer carry. A topic you’ve stopped following.

Forte sorts anything crossing his desk by asking four questions:

1. Does this move a current project forward?

2. Does it uphold an area?

3. Does it feed a resource?

4. If none of these apply, delete it.

That last option is the one people skip. Instead of deleting, they file, and Forte calls the habit procrastination in disguise. Filing feels productive, but it lets you postpone the task you’d rather not face.

Subscribe now

Projects End. Areas Don’t.

Projects end. Areas continue indefinitely. Everything sitting in front of you right now belongs to one or the other, and knowing which one changes how it feels to look at it.

Forte is upfront about where PARA breaks down. It collapses when the system gets as complicated as the life it’s meant to organize. If keeping it updated requires more attention than the tasks themselves, you’ll drop it the first hard week you have.

To keep the system from collapsing, Forte built two different modes into PARA itself, one for good days and one for hard ones.

On a good day, you zoom out to areas and resources.

On a hard day, when you’re only handling what’s urgent, you retreat straight into your Projects folder and work on one thing.

This is what keeps you organized on the days everything is falling apart, when a full weekly review is out of reach, and all you can manage is the next task in front of you.

Nearly every system people build only works on the good days. Yours can work on the bad ones too, without you having to keep it together by memory and willpower, once something else takes over the upkeep for you.

I’m publishing the next piece, ‘How to Build an AI Second Brain That Works on Your Worst Days, Not Just Your Best,’ this weekend on Between Thinking and Doing. It walks through an AI layer that sorts your notes into Projects, Areas, Resources, and Archives while you’re too buried to do it yourself, so the system keeps working even on the days you don’t.

See you then!

Subscribe now